Rethinking Ultrasound Augmentation: A Physics-Inspired Approach

Tirindelli, Maria Cristina; Eilers, Christine; Simson, Walter; Paschali, Magdalini; Azampour, Mohammad Farid; Navab, Nassir

doi:10.1007/978-3-030-87237-3_66

Cited by 17 publications

(6 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Concretely, one cannot gain information present in an MRI image with a CycleGAN transformed CT scan or vice versa 43 . Despite this knowledge, integrating PIPs into conditional CycleGANs 15 or Pseudo-physical augmentations 17 may improve model generalization despite not being fully rigorous and physically aware methods but rather “physically inspired” solutions.…”

Section: Discussionmentioning

confidence: 99%

“…Dataset enlargement BigAug 16 models degrade an average of 11% (Dice score change) from source to target domain, substantially better than conventional augmentation for medical segmentation tasks. In 17 the authors create new ultrasound physics inspired augmentations for segmentation and classification tasks. In 18 the authors use MRI physics augmentations for deep learning MRI reconstruction.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Physical imaging parameter variation drives domain shift

Kilim

Olar

Joó

et al. 2022

Sci Rep

View full text Add to dashboard Cite

Statistical learning algorithms strongly rely on an oversimplified assumption for optimal performance, that is, source (training) and target (testing) data are independent and identically distributed. Variation in human tissue, physician labeling and physical imaging parameters (PIPs) in the generative process, yield medical image datasets with statistics that render this central assumption false. When deploying models, new examples are often out of distribution with respect to training data, thus, training robust dependable and predictive models is still a challenge in medical imaging with significant accuracy drops common for deployed models. This statistical variation between training and testing data is referred to as domain shift (DS).To the best of our knowledge we provide the first empirical evidence that variation in PIPs between test and train medical image datasets is a significant driver of DS and model generalization error is correlated with this variance. We show significant covariate shift occurs due to a selection bias in sampling from a small area of PIP space for both inter and intra-hospital regimes. In order to show this, we control for population shift, prevalence shift, data selection biases and annotation biases to investigate the sole effect of the physical generation process on model generalization for a proxy task of age group estimation on a combined 44 k image mammogram dataset collected from five hospitals.We hypothesize that training data should be sampled evenly from PIP space to produce the most robust models and hope this study provides motivation to retain medical image generation metadata that is almost always discarded or redacted in open source datasets. This metadata measured with standard international units can provide a universal regularizing anchor between distributions generated across the world for all current and future imaging modalities.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Physical imaging parameter variation drives domain shift

Kilim

Olar

Joó

et al. 2022

Sci Rep

View full text Add to dashboard Cite

show abstract

“…This motivates the need for transformations that emulate the physics of medical image acquisition. Such specialized augmentations have been developed for computed tomography [19], magnetic resonance tomography [24] and ultrasound imaging [31]. Even though the physics behind OCTA image acquisition is well understood [27], data augmentation schemes specific to this modality have not been introduced to date.…”

Section: Related Workmentioning

confidence: 99%

Physiology-Based Simulation of the Retinal Vasculature Enables Annotation-Free Segmentation of OCT Angiographs

Menten

Paetzold

Dima

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Optical coherence tomography angiography (OCTA) can noninvasively image the eye's circulatory system. In order to reliably characterize the retinal vasculature, there is a need to automatically extract quantitative metrics from these images. The calculation of such biomarkers requires a precise semantic segmentation of the blood vessels. However, deep-learning-based methods for segmentation mostly rely on supervised training with voxel-level annotations, which are costly to obtain. In this work, we present a pipeline to synthesize large amounts of realistic OCTA images with intrinsically matching ground truth labels; thereby obviating the need for manual annotation of training data. Our proposed method is based on two novel components: 1) a physiology-based simulation that models the various retinal vascular plexuses and 2) a suite of physics-based image augmentations that emulate the OCTA image acquisition process including typical artifacts. In extensive benchmarking experiments, we demonstrate the utility of our synthetic data by successfully training retinal vessel segmentation algorithms. Encouraged by our method's competitive quantitative and superior qualitative performance, we believe that it constitutes a versatile tool to advance the quantitative analysis of OCTA images.

show abstract

“…We also introduce several augmentations handcrafted specifically to the physics of RF ultrasound. Ultrasound physicsinspired augmentations have been proposed previously for Bmode [46] but not RF. Our approach considers the decomposition of an RF line into envelope and instantaneous frequency.…”

Section: Data Augmentationsmentioning

confidence: 99%

Self-Supervised Learning with Limited Labeled Data for Prostate Cancer Detection in High Frequency Ultrasound

Wilson¹,

Mahdi²,

Jamzad³

et al. 2022

Preprint

View full text Add to dashboard Cite

Deep learning-based analysis of highfrequency, high-resolution micro-ultrasound data shows great promise for prostate cancer detection. Previous approaches to analysis of ultrasound data largely follow a supervised learning paradigm. Ground truth labels for ultrasound images used for training deep networks often include coarse annotations generated from the histopathological analysis of tissue samples obtained via biopsy. This creates inherent limitations on the availability and quality of labeled data, posing major challenges to the success of supervised learning methods. On the other hand, unlabeled prostate ultrasound data are more abundant. In this work, we successfully apply selfsupervised representation learning to micro-ultrasound data. Using ultrasound data from 1028 biopsy cores of 391 subjects obtained in two clinical centres, we demonstrate that feature representations learnt with this method can be used to classify cancer from non-cancer tissue, obtaining an AUROC score of 91% on an independent test set. To the best of our knowledge, this is the first successful end-to-end self-supervised learning approach for prostate cancer detection using ultrasound data. Our method outperforms baseline supervised learning approaches, generalizes well between different data centers, and scale well in performance as more unlabeled data are added, making it a promising approach for future research using large volumes of unlabeled data.

show abstract

Rethinking Ultrasound Augmentation: A Physics-Inspired Approach

Cited by 17 publications

References 20 publications

Physical imaging parameter variation drives domain shift

Physical imaging parameter variation drives domain shift

Physiology-Based Simulation of the Retinal Vasculature Enables Annotation-Free Segmentation of OCT Angiographs

Self-Supervised Learning with Limited Labeled Data for Prostate Cancer Detection in High Frequency Ultrasound

Contact Info

Product

Resources

About